* sd: remove C++ support for enforcing fixed LoRA multipliers
The logic at the Python level is enough.
* sd: support changing preloaded LoRA multipliers
We keep the same rules as before:
- Any LoRA with multiplier 0 can be changed
- If all LoRAs have multiplier != 0, they are fixed and optimized
but tweak the corner case of LoRAs specified more than once to
allow adjusting the multiplier if the same LoRA is also specified
with a zero multiplier, as if they were two different LoRAs.
So the following keeps working as before:
- --sdlora /loras/lcm.gguf --sdloramult 1 : fixed as 1
- --sdlora /loras/lcm.gguf --sdloramult 0 : dynamic, default 0
- --sdlora /loras/ : dynamic, default 0
- --sdlora /loras/lcm.gguf /loras/lcm.gguf --sdloramult 1 1 : fixed as 2
But now we have:
- --sdlora /loras/lcm.gguf /loras/lcm.gguf --sdloramult 1 0 : dynamic, default 1
- --sdlora /loras/lcm.gguf /loras/ --sdloramult 1 : dynamic, default 1
* backend support for controlling LoRA cache and fixed multipliers
The generation LoRA multipliers are now added to the initial
multipliers, so e.g. a merged LCM model will behave the same as
a normal model with a preloaded LCM LoRA.
* frontend support
* sd: sync to master-525-d6dd6d7
* sd: add support for cache modes for inference acceleration
* keep gendefaults as a JSON object inside the config file
* covered more invalid cases on gendefaults parsing
`0. in inputs.lora_multipliers` didn't work because the C array has
variable length.
Also fixed a few corner cases related to the default multipliers
(mainly to ensure robustness against future changes, since in most
cases the multiplier list is already sanitized by a previous
function).
* fix corner case in sd_oai_transform_params
Also fix typo in the function name.
* support for customizing loaded LoRA multipliers
The `sdloramult` flag now accepts a list of multipliers, one for each
LoRA. If all multipliers are non-zero, LoRAs load as before, with no extra
VRAM usage or performance impact.
If any LoRA has a multiplier of 0, we switch to `at_runtime` mode, and these
LoRAs will be available to multiplier changes via the `lora` sdapi field and
show up in the `sdapi/v1/loras` endpoint. All LoRAs are still preloaded on
startup, and cached to avoid file reloads.
If the list of multipliers is shorter than the list of LoRAs, the multiplier
list is extended with the first multiplier (1.0 by default), to keep it
compatible with the previous behavior.
* support for `<lora:name:multiplier>` prompt syntax and metadata
* add a few tests for sanitize_lora_multipliers
* common : fix Step-3.5-Flash format detection and thinking support
Step-3.5-Flash uses the same XML-style tool call format as Qwen3-Coder
(<tool_call><function=...><parameter=...>) but its Jinja template lacks
the bare <function> and plural <parameters> markers that the detection
logic previously required. This caused it to fall through to Hermes 2
Pro, which doesn't call func_args_not_string(), so arguments stayed as
JSON strings and templates using arguments|items crashed.
Additionally, the Qwen3-Coder-XML format handler had no thinking support.
Models like Step-3.5-Flash that unconditionally emit <think> in their
generation prompt need the same thinking_forced_open handling that
Nemotron v3 and Hermes 2 Pro already have, otherwise reasoning_content
is never separated from content in API responses.
Changes:
- Relax Qwen3-Coder XML detection to only require the 3 shared markers
- Tighten Nemotron v3 branch to also require bare <function> and plural
<parameters>, preventing Step-3.5-Flash from being misrouted via <think>
- Add thinking_forced_open support to Qwen3-Coder-XML init function
- Add <think>/</think> to preserved tokens
- Fix build_grammar_xml_tool_call to handle thinking_forced_open in the
grammar root rule, allowing </think> before tool calls
- Add Step-3.5-Flash chat template and format detection test
Builds on: https://github.com/ggml-org/llama.cpp/pull/19283
* chat : route Step-3.5-Flash to Nemotron v3 PEG parser, add tests
Step-3.5-Flash uses the same XML tool call format as Qwen3-Coder and
Nemotron 3 Nano (<tool_call>/<function=...>/<parameter=...>) but with
unconditional <think> output. Route it to the Nemotron v3 PEG parser
for streaming and schema-aware parameter parsing.
Detection: templates with <think> + XML tool tags use Nemotron v3 PEG
parser; templates without <think> (Qwen3-Coder) use GBNF grammar.
Tests cover: basic messages, tool calls with/without thinking content,
parallel tool calls, code string parameters, optional </parameter>
closing tags, and JSON schema response format.
* chat : remove dead thinking code from qwen3_coder_xml
Remove thinking handling code that became unreachable after routing
Step-3.5-Flash to the Nemotron v3 PEG parser. Qwen3-Coder has no
<think> in its template, so the thinking_forced_open logic, preserved
tokens, and grammar prefix were dead paths.
* fix vulkan ggml_acc only works in 3d but not 4d
* removed clamp in test_acc_block
* use the correct stride and its test case
* cuda : fix "supports op" condition
* change src0 to src1 in ggml_vk_acc. Update acc.comp with jeffbolznv\'s suggestion except to keep the boundary check
* version without boundary check
* revert back to boundary check version
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Write out a 2-bit code per block and avoid loading the mask when it
matches these two common cases.
Apply this optimization when the mask is relatively large (i.e. prompt
processing).
* jinja : add missing 'in' test to template engine (#19004)
The jinja template parser was missing the 'in' test from
global_builtins(), causing templates using reject("in", ...),
select("in", ...), or 'x is in(y)' to fail with
"selectattr: unknown test 'in'".
This broke tool-calling for Qwen3-Coder and any other model
whose chat template uses the 'in' test.
Added test_is_in supporting array, string, and object containment
checks, mirroring the existing 'in' operator logic in runtime.cpp.
Includes test cases for all three containment types plus
reject/select filter usage.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* reuse test_is_in in binary op
---------
Co-authored-by: Sid Mohan <sidmohan0@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
* undefined is treated as iterable (string/array) by filters
`tojson` is not a supported `undefined` filter
* add tests
* add sequence and iterable tests
keep it DRY and fix some types
* mla : pass V as a view of K to the FA op
* cuda : adjust mla logic to new layout
* kv-cache : fix rope shift
* tests : remove comment
* cuda : fix reusable_cutoff
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
